Jump to: navigation, search

Difference between revisions of "StarlingX/StarlingX Packet.com iPXE Installation"

(Using Layer 2 VLANs in Packet.com)
m (Run config_controller)
Line 57: Line 57:
=== Run config_controller ===
=== Run config_controller ===
==== Notes ====
* For the c1.large.x86 it seems like config_controller will fail without using the nvme drive, perhaps related to size of the disks.
== Using Layer 2 VLANs in Packet.com ==
== Using Layer 2 VLANs in Packet.com ==

Revision as of 16:32, 13 April 2019

Install AIO Simplex into Packet.com via iPXE

Packet.com is a baremetal public cloud, and they have donated some resources to the StarlingX project. The way that a custom operating system is installed into Packet.com is via iPXE. These instructions show a basic method for initial installation of a StarlingX ISO on Packet.com.

Configure a Web Server to serve ISO and iPXE Confguration

This assumes an Ubuntu 16.04 instance, but any Apache web server should do. It must be available publicly, ie. have a public IP address that is available from the Packet.com data center that the instance is being deployed to. (Typically this would be an instance that is running in the same Packet.com datacenter, but it doesn't have to be.)

Install Apache.

apt install apache -y

Download an ISO from the CENGN StarlingX build archive.

Mount that ISO where it will be available to the webserver process.

mkdir /var/www/html/stx
mount -o loop ~/bootimage.iso /var/www/html/stx

Create an iPXE configuration file that is available from the web server. Replace the "webserver_public_ip" with the webservers public IP address. The configuration below will install a AIO Simplex node via the kickstart file indicated in the kernel line. This configuration needs to be available on a public webserver, and would usually be installed on the same webserver as the ISO was mounted on.

NOTE: This is a configuration that is currently working, it may not be the perfect setup. Please feel free to make it better.

set base-url http://<webserver_public_ip>/stx
kernel ${base-url}/vmlinuz console=ttyS1,115200n8 root=live:${base-url}/LiveOS/squashfs.img ip=dhcp ks=${base-url}/smallsystem_ks.cfg boot_device=sda rootfs_device=sda inst.text inst.repo=${base-url} security_profile=standard user_namespace.enable=1
initrd ${base-url}/initrd.img

Using a Compute Type with nvme drives

Add the device name to the kernel line to the iPXE configuration file.

Example entry below:


Create an Instance in Packet.com

Use "Custom iPXE" for the operating system choice and point it to the iPXE configuration URL that was setup in the webserver configuration step.

The currently used node type is c1.small.x86.

Then access the Packet.com instance via it's "Out of Band Console." Once the instance is available and you can connect to the out of band console you should be able to see the instance booting from the iPXE configuration.

Once the installation has completed the node will reboot and will not boot by PXE again unless requested. At this point it seems the public IP that is provided by Packet.com is available.on the first interface and is accessible from the Internet via SSH.

Run config_controller



  • For the c1.large.x86 it seems like config_controller will fail without using the nvme drive, perhaps related to size of the disks.

Using Layer 2 VLANs in Packet.com

Packet, which by default is layer 3 only, also has the option to configure VLANs.

  • While the documentation suggests they have some kind of "internet gateway" functionality, it does not seem like that is accessible yet. Instead one currently would have to setup another node as a hybrid node to act as a gateway.
  • The CNCF appears to have some code around working with VLANS